Blue Cloud Softech Solutions has won a large-scale data annotation and AI training services contract from US-based Stratos Forge, following the completion of a paid pilot programme.
The new order has a stated commercial value of ₹110.08 crores. The pilot phase had an approximate value of ₹18.00 Crores.
Blue Cloud reported achieving an annotation accuracy of 96.68% in the pilot. The result was measured against predefined metrics agreed with Stratos Forge.
Stratos Forge is an AI-focused technology company based in New Jersey. It develops enterprise systems, digital automation tools, analytics products, and machine learning platforms for global customers.
Blue Cloud said it will execute the whole project through its existing delivery infrastructure. The company will also use its Centre of Excellence partnerships with universities as part of the engagement.
The project will focus on large-scale data annotation for AI training. The work will cover complex data types and advanced workflow models.
Shift in annotation
Data annotation has expanded beyond basic manual labelling. It now includes automation-driven workflows and support for 3D LiDAR, structured text and high-resolution imagery.
Blue Cloud plans to deploy its annotation ecosystem for Stratos Forge. The ecosystem combines automation, human review, and structured quality controls.
The company will apply AI-assisted, automated annotation methods to the account. These include active learning, pre-labelling, weak supervision and synthetic data generation.
Active learning models will identify ambiguous or low-confidence samples. Human reviewers will then handle these selected items.
Pre-labelling will use pre-trained models for initial annotations. Human staff will correct and refine the results.
Weak supervision will create labels using code-based functions. This method will apply in areas such as text and classification workloads.
Synthetic data generation will use techniques such as GANs and diffusion models. These techniques will produce labelled datasets that support privacy requirements and volume demands.
Quality controls
Blue Cloud also plans to apply a structured quality control framework to the project. The framework will include inter-annotator agreement checks, gold sets and semantic consistency rules.
Inter-annotator agreement will involve multiple annotators labelling the same sample. Statistical models such as majority vote and Dawid-Skene will then infer the final label.
The company will measure consistency using metrics such as Cohen's Kappa. These metrics will indicate the level of agreement between annotators.
Gold sets, also known as sentinel data, will serve as hidden ground-truth samples. The system will use these samples to track annotator accuracy in real time.
Automated checks will flag or reject low-accuracy work. Review processes will then examine these samples.
Semantic consistency checks will enforce logical rules on labels. These rules will block invalid label combinations from entering training datasets.
Human-in-the-loop
The project will use a human-in-the-loop workflow. The model will predict outputs, and human reviewers will correct errors. The corrected data will support further model retraining.
Blue Cloud will also use micro-tasking methods. It will break complex annotation work into smaller, focused steps.
The company expects this structure to reduce annotator bias and cognitive load. It also expects more precise labels from the narrower tasks.
Domain experts within Blue Cloud will contribute to the work. Academic partners in its Centres of Excellence will also contribute.
The firm plans to apply specialised annotation approaches across several domains. These include autonomous systems, robotics and industrial vision, natural language processing, knowledge engineering, behavioural analytics, and 3D LiDAR and point-cloud mapping.
Customer relationship
Blue Cloud framed the new contract as an expansion of its existing relationship with Stratos Forge. The companies first worked together on the now-completed pilot programme.
"We are delighted to expand our partnership with Stratos Forge Inc after our highly successful pilot engagement. Their confidence in BCSSL highlights the strength of our annotation automation frameworks, our CoE talent pipeline, and our ability to deliver world-class AI training data at scale," said Janaki Yarlagadda, Chairman, Blue Cloud Softech Solutions.
